The Longer – 2 of 2
Perhaps the phrase ‘homeless count’ causes people to assume that the purpose – and therefore design – of the homeless count is to count all the homeless and it is that erroneous assumption that leads to the utterance of utter nonsense such as:
“Earlier this year, a homeless count found a dramatic increase in homelessness across the region from three years ago, including a 79 per cent increase in Abbotsford.
A moment’s thought and it is obvious that a count to achieve an accurate reflection of the number of homeless – within an acceptable margin of error* – is not going to be accomplished using the current method or within the current timeframe. A count to provide – within an acceptable margin of error – an accurate reflection of the number of homeless would be more complex and require a longer timeframe.
*given that the nature of the population being counted the probability of 100% accuracy in the count approaches zero, therefore the method the count is conducted in MUST focus on controlling the magnitude of error [reducing and/or controlling variables that induce error]. Only by holding the counts within a small margin of error will the counts provide results that are comparable between different counts and different years
The purpose of the count has never been to get an accurate count of the actual number of homeless.
The count’s purpose is to be able to compare the number of homeless between different years in order to draw conclusions about what is happening with, and within, the homeless population.
“In our lust for measurement, we frequently measure that which we can rather than that which we wish to measure… and forget that there is a difference.” George Udny Yule
If the comparison of the results of homeless counts in different years are to be anything other than misleading gibberish one must eliminate or control all the variables that can move the number of homeless counted up or down.
Weather is an example of a variable that can have a pronounced effect on the number of homeless counted and over which the count has no control. Pouring rain instead of sun will significantly change the behaviour of the homeless [and the counters] and therefore will significantly change the number of homeless counted.
And weather is only one of numerous variables that can distort the number of homeless counted.
Simple high school math: you cannot compare apples and oranges.
Therefore, for a comparison between the results of the 2014 homeless count and the results of the 2017 count to have been meaningful, rather than the misleading gibberish they are, it is necessary that the counts be reduced to comparable units so one is not comparing apples to apples
The first requirement in using statistics is that the facts treated shall be reduced to comparable units. Claude Bernard
A simple application of algebra [2014] + [2014]X @ [2017 comparable units] will transmute the homeless totals into comparable units; transmute oranges into apples so one is comparing apples to apples not apples to oranges.
X is the adjustment needed to adjust the 2014 homeless count total to be the same magnitude [percentage of the actual total number of homeless] as the 2017 homeless count total.
The fact the homeless count does not control or compensate for the variables that effect [raise and/or lower] the number of homeless counted, means that for any given year the counts deviation from the actual number of homeless is unknown. It also means the deviation [degree of inaccuracy] varies from count to count, depending on the effect the variables have on the count in differing years.
The fantasy generated in Abbotsford by comparing the results of the 2017 count with the 2014 count provides a vibrant example of the distortion and misinformation that result from ignoring basic mathematics and comparing apples to oranges
In order to limit the complexity** of the example it will be limited to examining the effect of the count taking place on the last extreme weather protocol day of the 2016/17 winter.
** The method and timing of the homeless count ensures the total number of homeless counted is affected by an unknown number of variables that have unknown [variable] effects such that the count result contains an unknown but significant error in the count’s number of homeless counted versus the actual number of homeless in the count area
The bitter cold and snow of the winter of 2016/17 resulted in extra spaces for the homeless to shelter and motivated the homeless to find warmth.
The 2017 homeless count being conducted on the last extreme weather day of the 2016/17 extreme weather season resulted in a significantly higher percentage of the homeless population being accessible/available to be counted.
The effect of the extreme weather was to increase the number of homeless counted by 50% over the number who would have been counted if the homeless count had take place on a non-extreme weather protocol day. With extreme weather making more of the homeless accessible and counted the count counted 271 homeless1.
Consulting, at the time the count occurred, with those working with the homeless population and who have paid attention to and discussed the growth in the homeless population year over year over the last decade places the effect of extreme weather on the number of homeless counted at 50%..
In other words, if the 2017 count had not been significantly increased by being on an extreme weather day the total number of homeless counted would have been 180. [270 – 90 = 180; 90 + 90 + 90 = 270; 90/180 = 50%.]
If you compare 2014’s total of 141 to the adjusted 2017 total of 180 versus the comparison between 2014’s 141 and 2017’s unadjusted 271 total it is clear 1) the failure to control variables gives rise to count results from different years that are not comparable 2) comparing count numbers from different years has a propensity to mislead and result in erroneous conclusions and false knowledge.
Knowing the effect of extreme weather on the count [50% increase] you could adjust 2014 results instead of the 2017 results.
[2014] + [2014]X @ [2017 comparable units]; [141] +[141].5 = [2017 CU];
141 +71 = [2017 CU]; [2017 CU] = 212;
Note the significant difference between adjusting 2017 [170 – 141 = 29] and adjusting 2014 [271 – 212 = 59].
The effect of the variables is such that if you were to conduct a homeless count on one day and then conduct the homeless count again the next day, you would get different results. The results of the two counts could be reasonable close, notably higher or notably lower depending on the nature of the variables affecting [or not affecting] the two counts. Even if the results from the two counts are reasonably close the number of homeless counted on one day would contain people not counted on the next day. Ironically the apparent closeness of the counts is in fact a result of uncontrolled variables that have different people being counted one day and missed on the second day.
Another entertaining irony is that in order to adjust the homeless results you need to know the actual number of homeless in the count area. Of course if you know the number of homeless you have no need to make any adjustment.
Counting the homeless and being able to compare the results with the results in prior years requires designing and conducting a count that controls and /or offsets the variables affecting the count.
No matter how hard you ignore the reality that comparisons between different counts and conclusions based on those comparisons are meaningless gibberish, you cannot make the comparisons and conclusions one iota less gibberish.
“You can’t fix by analysis what you bungled by design.” Light, Singer and Willett, page v
Given how glaringly blatant and obvious it is that comparisons and conclusions are nothing but meaningless gibberish, the question for those responsible for misleading the pubic is Why?
Just as important is why [and how] is it that City Council chooses to be so unaware, so wilfully ignorant of the nature and state of the homeless population in Abbotsford that they swallow such obvious bunkum as Truth?